The AI market is loud right now. Every software company is leading with AI, every conference has an AI track, and most healthcare organizations are somewhere on the spectrum between cautiously curious and quietly overwhelmed. Feel familiar?
What’s getting lost in the noise is a crucial question: is the AI your organization uses built for healthcare? Or, is your team trying to work with a tool that wasn’t built for your day-to-day reality?
That's the conversation our Chief Sales Officer, Dale Van Gorder, and Chief Product Officer, Lily He, sat down to have this week. It wasn’t a product demo or a feature overview. It was a data-driven look at what healthcare-specific AI means, why the distinction matters and how to evaluate whether your current tools meet that standard.
There are four meaningfully different types of AI in common use today. They carry different risk profiles, produce different kinds of outputs and behave very differently when asked to reason about complex healthcare questions.
The most counterintuitive finding from recent research: fine-tuning an AI model on medical data doesn't necessarily make it safer. In some cases it makes it less safe. That's the kind of finding that doesn't show up in a vendor pitch, but it absolutely should show up in your evaluation process.
Most people evaluate AI on whether it gets the right answer. But there's a second question that matters just as much in healthcare: does it get the same answer consistently?
Ask a general-purpose AI the same question three times. You'll likely get three different answers. In a compliance or contract management context, that's not just inconvenient. It creates real risk, especially when those outputs are feeding into reports, workflows or decisions that people are relying on.
One of the most useful frameworks from the session involves what happens when AI touches your data more than once. If accuracy starts at 80% and data passes through three AI steps, which is very common in a compliance workflow, you're not operating at 80% accuracy anymore.
The output still looks credible though, and that's the part that matters most.
The regulatory environment around AI in healthcare is moving faster than most organizations realize. State legislatures are acting. Federal agencies are watching. And the standard they're setting centers on one thing: human oversight.
Organizations that are building that oversight into their AI workflows now aren't just doing the responsible thing. They're getting ahead of where compliance requirements are heading.
The session closed with five questions every healthcare organization should be asking about any AI tool they use, including ours. They're simple questions. Most vendors won't love them, and the answers will tell you more than any demo will.